Side_1_360

4 Controlling Voice Quality

The conclusion from Section 3 is that for our

purposes the rating Rcan be written as

R= Rint,G.711– Id(TM 2 E,m, EL 1 , EL 2 ) –

Ie(codec,Ploss) (10)

The combined effect of the first and second term

is illustrated in Figure 3. The third term is dis-

played in Figure 5.

4.1 Quality Bounds

Since the echo control bound of 25 ms is almost

always exceeded when the phone calls are trans-

ported over a packet-based network, echo con-

trol is strongly recommended for packetized

phone calls. A good echo controller, i.e. an echo

canceller compliant with ITU-T recommenda-

tion G.168 [3], can increase the echo loss (from

20 dB usually occurring in the PSTN) to 50 dB.

With an echo controller with a non-linear ele-

ment perfect echo control, in which case the

echo loss is increased to infinity, can be

achieved.

From Figure 3 it is clear that in the case of per-

fect echo control at both sides, the intrinsic qual-

ity of the call is attained if the mean one-way

mouth-to-ear delay is kept below 150 ms. From

eq. (10) we notice that this intrinsic quality is

solely determined by the distortion impairment

Ie, which in turn is determined by the codec(s)

used and the overall packet loss experienced.

Since the intrinsic quality Rint,G.711of an undis-

torted call is about 94 and the bound for tradi-

tional quality is 70, there is an impairment bud-

get of 24, part of which is consumed by the

codec(s) (see Figure 4). Once the codec has been

chosen, the remainder of the margin can be con-

sumed either by allowing the mean one-way

mouth-to-ear delay to exceed 150 ms or by toler-

ating some packet loss. The bound on the mean

one-way mouth-to-ear delay for a certain codec

is derived by subtracting the impairment associ-

ated with that codec (displayed in Figure 4) from

the curves of Figure 3 and determining where

the curve associated with perfect echo control

drops below 70. The bound on packet loss for a

certain codec is derived by determining in Fig-

ure 5 for which packet loss value the impairment

budget of 24 is just not consumed. The bounds

for the AMR codec are derived under the

assumption that the interpolation (i.e. the curve

in Figure 4 labeled “AMR”) is valid and that per

percent packet loss 4 units are added to the

impairment Ie. The fourth column of Table 3

and Table 4 gives the codec-dependent bounds

on the mean one-way mouth-to-ear delay and

packet loss, respectively, when the echo is per-

fectly controlled [10].

Table 3 Tolerable mean one-way mouth-to-ear delay TM2Ebounds when there is no
packet loss in the case of perfect echo control (EL=inf) and an echo loss
EL = 50 dB. (NA = Traditional PSTN quality (R = 70) is Not Attainable.)

Standard Short name Codec bit TM 2 E(ms) TM 2 E(ms)

body rate (kb/s) EL= infinite EL= 50 dB

ETSI GSM-HR 5.6 177 29

ITU-T GSM-FR 13 21 106

ITU-T G.711 64 400 291

ITU-T G.728 12.8 210 106

16 322 243

ITU-T G.726 16 NA NA

G.727 24 NA NA

32 322 243

40 375 276

ITU-T G.723.1 5.3 219 131

6.3 251 187

ITU-T G.729 8 294 223

ETSI GSM-EFR 12.2 342 256

3GPP AMR 4.75 197 72

5.15 214 117

5.9 239 172

6.7 262 198

7.4 280 213

7.95 293 223

10.2 332 250

12.2 342 256

Table 4 Tolerable packet loss Plossbounds for a mean one-way mouth-to-ear delay

below 150 ms in the case of perfect echo control and for a mean one-way mouth-to-

ear delay of 150 ms for an echo loss EL = 50 dB. (NA = Traditional PSTN quality

(R = 70) is Not Attainable.)

Standard Short name Codec bit Ploss Ploss

body rate (kb/s) EL= infinite EL= 50 dB

TM 2 E< 150 ms TM 2 E= 150 ms

ITU-T G.711 no PLC 64 1.2 0.9

ITU-T G.711 PLC 6.4 9.6 6.1

ITU-T G.723.1 6.3 2.1 0.6

ITU-T G.729 8 3.5 1.7

ETSI GSM-EFR 12.2 2.5 1.5

3GPP AMR 4.75 0.7 NA

5.15 1.2 NA

5.9 1.9 0.3

6.7 2.6 1.1

7.4 3.2 1.6

7.95 3.5 2.0

10.2 4.6 3.0

12.2 4.8 3.2